An automatic detection method for surface defects in printed materials using machine vision

Abstract

This invention discloses an automatic detection method for printed surface defects using machine vision, belonging to the field of printing defect detection technology. The method includes the following steps: synchronously collecting data during the printing process to form a multimodal time-series dataset; extracting and quantifying key dynamic features from the multimodal time-series dataset to generate a unified system state vector; constructing a system state stability evolution model, inputting the system state vector into the stability evolution model, calculating and outputting a real-time system stability index; responding to a real-time system stability index falling below a preset warning threshold, initiating causal decoupling analysis to determine the current dominant failure path; and generating differentiated intervention instructions based on the dominant failure path. This invention achieves a transformation from passive fault response to proactive trend prediction, from fuzzy fault diagnosis to precise failure attribution, and from universal intervention to differentiated intelligent control, ensuring product quality and improving production efficiency without interrupting production.

Description

Technical Field

[0001] This invention relates to the field of printing defect detection technology, and more specifically to an automatic detection method for surface defects in printed materials using machine vision. Background Technology

[0002] High-speed roll-to-roll printing is crucial for large-scale, high-efficiency production, and its product quality places high demands on system stability. During the printing process, the evaporation rate of ink solvents affects product drying and spreading, while the micro-vibrations generated by the high-speed operation of the equipment affect ink transfer and substrate transport.

[0003] Existing technologies typically employ independent monitoring systems. These systems are not only slow to react, only triggering alarms after defects have occurred on a large scale, which can easily lead to a large number of scraps, but also make it difficult to quickly and accurately determine the root cause of the problem. In addition, they suffer from problems such as limited intervention strategies, low efficiency, potential introduction of new disturbances, and inability to effectively identify and decouple critical transitions in the system triggered by chemical and physical factors. Summary of the Invention

[0004] The purpose of this invention is to provide an automatic detection method for surface defects of printed materials using machine vision, which solves the problems existing in the background art.

[0005] To address the aforementioned technical problems, this invention provides an automatic detection method for surface defects in printed materials using machine vision, comprising the following steps:

[0006] S1. Synchronously collect time-series defect image data, key component vibration data, and substrate surface temperature data during the printing process to form a multimodal time-series dataset;

[0007] S2. Extract and quantify key dynamic features from the multimodal time-series dataset to generate a unified system state vector; the key dynamic features include visual features, vibrational features, and thermodynamic features;

[0008] S3. Construct a system state stability evolution model, input the system state vector into the stability evolution model, calculate and output the real-time system stability index; the stability evolution model is constructed based on the deviation between the system state vector and the preset benchmark stable state vector;

[0009] S4. In response to the real-time system stability index falling below a preset warning threshold, initiate causal decoupling analysis, calculate the chemical dominant contribution and physical dominant contribution based on the system state vector, and determine the current dominant failure path by comparing the chemical dominant contribution and physical dominant contribution.

[0010] S5. Based on the dominant failure path, generate differentiated intervention instructions; if the dominant failure path is determined to be chemically dominant, generate adjustment instructions for the ink mixing system; if the dominant failure path is determined to be physically dominant, generate early warning instructions for the equipment maintenance system.

[0011] Preferably, S2 specifically includes:

[0012] S21. Based on the temporal defect image data, visual feature sub-vectors are extracted and combined by calculating image blur, texture entropy and color gradient norm.

[0013] S22. Based on the vibration data of key components, the energy amplitude at the relevant characteristic frequencies of the key components is extracted by performing a short-time Fourier transform on the vibration data of the key components, and combined into a vibration feature sub-vector.

[0014] S23. Based on the surface temperature data of the substrate, the average surface temperature, temperature standard deviation and maximum surface temperature gradient are calculated, and thermodynamic feature sub-vectors are extracted and combined.

[0015] S24. Normalize and merge the visual feature sub-vector, the vibration feature sub-vector, and the thermodynamic feature sub-vector into the unified system state vector.

[0016] Preferably, S3 specifically includes:

[0017] S31. Based on the system state vector and the preset benchmark stable state vector, calculate the difference between the two to obtain the system state deviation vector;

[0018] S32. By performing a weighted quadratic form calculation on the system state deviation vector using a preset weight matrix, a system potential energy value is generated that characterizes the degree to which the system deviates from the stable point in the potential energy field.

[0019] S33. Based on the time series of the system potential energy value, the real-time system stability index is obtained by calculating the product of the autocorrelation and variance of the time series of the system potential energy value and taking the reciprocal of the product.

[0020] Preferably, the reference steady-state vector is obtained by continuously collecting and analyzing printing process data with a preset yield rate higher than a specific standard, and calculating the time average of the calculated system state vector at each moment.

[0021] The preset weight matrix is ​​optimized by using Lasso regression analysis to analyze historical fault data and quantify the contribution of each key dynamic feature to the final product quality defects.

[0022] Preferably, S4 specifically includes:

[0023] S41. Decouple the system state vector and the preset weight matrix into a chemical part and a physical part, respectively. The chemical part corresponds to the visual and thermodynamic characteristics related to ink, and the physical part corresponds to the vibration characteristics related to equipment.

[0024] S42. Calculate the chemical dominant contribution based on the system state vector, baseline stable state vector, and weight matrix of the chemical component;

[0025] S43. Calculate the physical dominant contribution based on the system state vector, baseline stable state vector, and weight matrix of the physical component;

[0026] S44. Calculate the ratio of the chemical dominant contribution to the physical dominant contribution to obtain the dominant pathway ratio;

[0027] S45. If the dominant pathway ratio is higher than a preset chemical dominance decision threshold, the dominant failure pathway is determined to be chemically dominant; if the dominant pathway ratio is lower than a preset physical dominance decision threshold, the dominant failure pathway is determined to be physically dominant.

[0028] Preferably, the chemically dominant decision threshold and the physically dominant decision threshold are obtained through controlled fault injection experiments, specifically including:

[0029] a) By introducing a disturbance of excessively rapid solvent evaporation into the ink system during the experiment, the statistical distribution of the dominant pathway ratio under this condition was continuously recorded and analyzed to determine the chemical dominance decision threshold;

[0030] b) By applying known weak mechanical vibration excitation to key rotating components in experiments, continuously recording and analyzing the statistical distribution of the dominant pathway ratio under this condition, the physical dominance decision threshold can be determined.

[0031] Preferably, the preset warning threshold is determined by statistically analyzing a large amount of system data transitioning from a normal state to a fault state, and based on receiver operation characteristic curve analysis, selecting a corresponding value that maximizes the warning recall rate while ensuring an acceptable false positive rate.

[0032] Preferably, S5 specifically includes:

[0033] S51. If the dominant failure pathway is determined to be chemically dominant, an instruction will be automatically generated and sent to the ink viscosity control system to adjust the solvent replenishment parameters, and at the same time, an abnormal ink status prompt message will be pushed to the human-machine interface.

[0034] S52. If the dominant failure path is determined to be physically dominant, then based on the characteristic frequency information in the vibration characteristic sub-vector, the specific abnormal vibration component is identified, and a high-priority preventive maintenance work order for that component is automatically generated.

[0035] An automatic detection system for surface defects in printed materials using machine vision is also provided, comprising:

[0036] The multimodal data synchronous acquisition module is used to synchronously acquire temporal defect image data, key component vibration data, and substrate surface temperature data during the printing process to form a multimodal time-series dataset.

[0037] The dynamic feature vector construction module is used to extract and quantify key dynamic features reflecting the system state from the multimodal time series dataset and fuse them into a unified system state vector.

[0038] The system state stability evolution modeling module is used to construct a model describing the dynamic evolution of the system state vector based on the system state vector, and to calculate and output the real-time system stability index based on the model.

[0039] The dominant failure path causal decoupling module is activated in response to a real-time system stability index falling below a preset warning threshold to analyze the system state vector, thereby decoupling and determining the chemically or physically dominant failure path that causes system instability.

[0040] The precise differentiated intervention decision module is used to generate and output differentiated control instructions for chemical or physical pathways based on the dominant failure pathway determined by the causal decoupling module.

[0041] Beneficial effects

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. By constructing an evolutionary model that can proactively assess system stability and uniquely decoupling the root causes of system instability into chemically-dominated and physically-dominated failure pathways, this method achieves a transformation from passive fault response to proactive trend prediction, from fuzzy fault diagnosis to precise failure attribution, and from universal intervention to differentiated intelligent regulation. Through the stability index, predictive proactive intervention is achieved, identifying instability trends before quality declines and solving the problem of lag response.

[0044] 2. By generating differentiated intervention instructions based on diagnostic results, production efficiency can be improved, waste and unnecessary downtime can be reduced, and the problem of a single intervention strategy can be solved. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a flowchart of the method of the present invention.

[0047] Figure 2 This is a logic block diagram of the system of the present invention. Detailed Implementation

[0048] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0049] Example 1:

[0050] Please see Figure 1 This invention provides an automatic detection method for printed surface defects using machine vision, comprising the following steps: S1, synchronously acquiring temporal defect image data, key component vibration data, and substrate surface temperature data during the printing process to form a multimodal time-series dataset; S2, extracting and quantifying key dynamic features from the multimodal time-series dataset to generate a unified system state vector, wherein the key dynamic features include visual features, vibration features, and thermodynamic features; S3, constructing a system state stability evolution model, inputting the system state vector into the stability evolution model, calculating and outputting a real-time system stability index, wherein the stability evolution model... The model is constructed based on the deviation between the system state vector and the preset benchmark stable state vector; S4, in response to the real-time system stability index being lower than the preset warning threshold, causal decoupling analysis is initiated, and the chemical dominant contribution and physical dominant contribution are calculated based on the system state vector. By comparing the chemical dominant contribution and the physical dominant contribution, the current dominant failure path is determined; S5, based on the dominant failure path, differentiated intervention instructions are generated; if the dominant failure path is determined to be chemically dominant, an adjustment instruction for the ink mixing system is generated; if the dominant failure path is determined to be physically dominant, a warning instruction for the equipment maintenance system is generated.

[0051] This embodiment provides an automatic detection method for surface defects of printed materials applied to a high-speed roll-to-roll gravure printing production line. In step S1, a high-precision clock source is used to time-stamp and synchronize the machine vision camera deployed at the quality inspection station, the accelerometer sensor installed on key load-bearing components such as the impression cylinder and doctor blade support, and the infrared thermal imager array installed at the drying oven outlet. This ensures that the collected time-series defect image data, key component vibration data, and substrate surface temperature data are strictly aligned in time series, forming a multimodal time-series dataset. Subsequent steps S2 to S5 are based on this unified dataset for analysis, aiming to overcome the common limitations of existing technologies. Addressing the technical shortcomings of delayed response and ambiguous causes, this method constructs an evolutionary model capable of proactively assessing system stability. It uniquely decouples the root causes of system instability into chemically-dominated and physically-dominated failure pathways. This approach achieves a transformation from passive fault response to proactive trend prediction, from fuzzy fault diagnosis to precise failure attribution, and from universal intervention to differentiated intelligent control. Its ultimate technical objective is to identify system instability trends before large-scale printing defects appear and automatically generate clearly directional adjustment or maintenance instructions, thereby ensuring product quality without interrupting production and significantly improving production efficiency and equipment health management.

[0052] Example 2:

[0053] According to the aforementioned method, S2 specifically includes: S21, based on the temporal defect image data, extracting and combining visual feature sub-vectors by calculating image blur, texture entropy, and color gradient norm; S22, based on the vibration data of key components, extracting the energy amplitude at relevant characteristic frequencies of key components by performing short-time Fourier transform, and combining them into vibration feature sub-vectors; S23, based on the surface temperature data of the substrate, extracting and combining thermodynamic feature sub-vectors by calculating the average surface temperature, temperature standard deviation, and maximum surface temperature gradient; S24, normalizing and fusing the visual feature sub-vectors, the vibration feature sub-vectors, and the thermodynamic feature sub-vectors into the unified system state vector;

[0054] The core function of the system state vector construction in this embodiment is to transform the multi-source heterogeneous data extracted in steps S21 to S23 into a quantifiable, normalized, and dimensionally unified system state vector through step S24. This provides a precise digital foundation for subsequent stability modeling and causal decoupling; firstly, based on temporal defective image data... Extract visual feature subvectors that reflect the fluid properties and spreading state of the ink. Its included image blur Characterizing ink leveling properties and texture entropy Characterizing the microscopic uniformity of the ink layer, color gradient norm This quantifies the sharpness of color edges; these three factors together constitute an indirect observation of chemical processes such as ink solvent evaporation. Secondly, based on vibration data of key components... By using short-time Fourier transform analysis, the inherent characteristic frequencies of the equipment are extracted. Energy amplitude at These amplitudes constitute the vibration characteristic subvectors that reflect the mechanical health of the equipment. Secondly, from the surface temperature data of the printing substrate In this study, thermodynamic feature vectors closely related to ink drying efficiency and energy transfer processes were extracted. It includes the average surface temperature Temperature standard deviation and the maximum surface temperature gradient This directly reflects the efficiency and uniformity of the drying process; finally, after processing using normalization methods such as Z-score standardization, these three feature sub-vectors are... , and Merge into a global high-dimensional system state vector:

[0055] ;

[0056] Indicates time A vector describing the overall state of the entire printing system;

[0057] express The visual feature subvector at time step is composed of image blur, texture entropy, and color gradient norm;

[0058] express The vibration characteristic sub-vector at a given moment is composed of the vibration amplitude at the characteristic frequency of the key component;

[0059] express The thermodynamic characteristic vector at time t is composed of the average surface temperature of the substrate, the standard deviation of temperature, and the maximum surface temperature gradient.

[0060] Represents the transpose operation of a vector;

[0061] This vector at each time It provides a comprehensive and accurate description of the overall operational status of the entire printing system.

[0062] Example 3:

[0063] According to the aforementioned method, step S3 specifically includes: S31, calculating the difference between the system state vector and the preset benchmark stable state vector to obtain the system state deviation vector; S32, generating a system potential energy value characterizing the degree of deviation of the system from the stable point in the potential energy field by performing a weighted quadratic form calculation on the system state deviation vector using a preset weight matrix; S33, obtaining the real-time system stability index by calculating the product of the autocorrelation and variance of the system potential energy value based on the time series of the system potential energy value, and taking the reciprocal of the product.

[0064] According to the aforementioned method, the baseline steady-state vector is obtained by continuously collecting and analyzing printing process data with a preset yield rate higher than a specific standard, and calculating the time average of the calculated system state vector at each moment; the preset weight matrix is ​​optimized by using Lasso regression analysis to analyze historical fault data and quantify the contribution of each key dynamic feature to the final product quality defects.

[0065] The system state stability evolution modeling in this embodiment aims to achieve early warning of critical transitions in the system; its core is the introduction of the system potential energy function and the real-time system stability index. ;

[0066] System potential energy function:

[0067] ;

[0068] Indicates the current moment The dimension obtained through multimodal data fusion is The real-time system state vector;

[0069] Representing a dimension as The baseline stable state vector represents the system's most ideal operating state;

[0070] Represent a The diagonal weight matrix, whose diagonal elements Representing the The weight of the influence of each state feature on system stability; This represents the system state deviation vector, which is the result of the calculation in step S31;

[0071] This represents the system's potential energy value, which is a scalar.

[0072] The design of this formula is inspired by the potential energy landscape theory in physics. A healthy printing system is considered to be at the bottom of the potential energy field and in a stable state. When the system is subjected to internal and external disturbances and begins to deviate from its optimal operating state, it is like a physical particle being pushed to a position with higher potential energy, and its potential energy value will increase. This function aims to build such a mathematical model to accurately quantify the degree to which the state of the entire high-dimensional system deviates from its ideal stable point with a single scalar, providing a clear and measurable basis for subsequent stability assessment.

[0073] In the real-time monitoring process, the system will collect and construct real-time state vectors. Substituting into the formula, the time series of potential energy values ​​can be continuously calculated. ;like If the value of continuously increases, it clearly indicates that the system is evolving from a stable state to an unstable critical state. This function condenses multi-dimensional and complex characteristic changes into a single index with clear physical meaning, making the judgment of the evolution trend of the overall system state intuitive and quantitative.

[0074] Select data from one or more production batches historically identified as having a product quality excellence rate exceeding a specific standard (e.g., 99%); continuously collect and calculate the system state vector at each moment during the operation of these batches. Finally, the arithmetic mean over time is calculated for all collected vectors, and this average vector is then determined. ;

[0075] A large amount of historical production data was collected, including system state vectors at various points in time and the corresponding final product defect rates. Lasso regression was used for analysis, with each feature of the system state vector as the input variable and the product defect rate as the target variable. Lasso regression's feature selection capability allowed the acquisition of the contribution coefficient of each feature to the defect rate. These coefficients were then normalized and used as the weight matrix. Weight values ​​on the diagonal This makes the potential energy function more sensitive to state changes that have historically been shown to be highly correlated with product defects.

[0076] Real-time system stability index :

[0077] ;

[0078] This represents the system potential energy value calculated from the aforementioned potential energy function;

[0079] The autocorrelation coefficient of this time series under a first-order delay is a dimensionless scalar.

[0080] This represents the variance of the time series within the same time window; step S33 is the complete calculation process of the index.

[0081] The theoretical basis of this index is the critical slowdown phenomenon in complex systems theory; when a dynamic system approaches its critical transition point, its speed of recovery to equilibrium after a small disturbance decreases significantly. This decrease in recovery ability manifests as a simultaneous and sharp increase in autocorrelation and variance in the time series of the system's state variables. Only the potential energy value is monitored. The increase or decrease can only determine the magnitude of the deviation, while The aim is to provide an earlier and more sensitive warning signal of critical transition by capturing the dynamic characteristic of a slowdown in system recovery speed;

[0082] During system operation, calculations are performed in real time based on a sliding time window. The first-order autocorrelation coefficient and variance of the time series are then obtained. Continuous values; during stable system operation, Its fluctuations are small and its recovery is fast; its autocorrelation and variance are both at low levels, making it... The index remained stable at a high level; when the system began to destabilize, the critical slowdown phenomenon led to... and As the synchronization increases significantly, its product also increases dramatically, causing the reciprocal of it to... The index dropped sharply; therefore, an early warning threshold was set and monitoring was conducted. Whether the value falls below this threshold allows for highly sensitive prediction of large-scale quality problems before they occur.

[0083] Example 4:

[0084] According to the aforementioned method, step S4 specifically includes: S41, decoupling the system state vector and the preset weight matrix into a chemical part and a physical part, respectively, wherein the chemical part corresponds to the visual and thermodynamic characteristics related to ink, and the physical part corresponds to the vibration characteristics related to equipment; S42, calculating the chemical dominance contribution based on the system state vector, the baseline stable state vector, and the weight matrix of the chemical part; S43, calculating the physical dominance contribution based on the system state vector, the baseline stable state vector, and the weight matrix of the physical part; S44, calculating the ratio of the chemical dominance contribution to the physical dominance contribution to obtain the dominance pathway ratio; S45, if the dominance pathway ratio is higher than a preset chemical dominance decision threshold, then the dominance failure pathway is determined to be chemically dominant; if the dominance pathway ratio is lower than a preset physical dominance decision threshold, then the dominance failure pathway is determined to be physically dominant.

[0085] According to the aforementioned method, the chemically dominant decision threshold and the physically dominant decision threshold are calibrated through controlled fault injection experiments. Specifically, this includes: introducing a disturbance that causes excessively rapid solvent evaporation into the ink system during the experiment, continuously recording and analyzing the statistical distribution of the dominant pathway ratio under this condition to determine the chemically dominant decision threshold; and applying known weak mechanical vibration excitation to key rotating components during the experiment, continuously recording and analyzing the statistical distribution of the dominant pathway ratio under this condition to determine the physically dominant decision threshold.

[0086] According to the aforementioned method, the preset warning threshold is determined by statistically analyzing a large amount of system data transitioning from a normal state to a fault state, and based on the receiver operation characteristic curve analysis, selecting the corresponding value that maximizes the warning recall rate while ensuring an acceptable false positive rate.

[0087] The causal decoupling analysis and decision-making in this embodiment aims to automatically and quantitatively diagnose the main driving factors leading to instability after the system issues an instability warning, thereby achieving precise intervention.

[0088] Causal contribution and pathway ratio:

[0089] ;

[0090] ;

[0091] ;

[0092] The chemical state subvector, composed of visual and thermodynamic features, is the result of decoupling in step S41.

[0093] This represents the physical state sub-vector composed of vibration characteristics, and is also the result of decoupling in step S41;

[0094] and These represent the reference stable state vectors respectively. The corresponding chemical and physical sub-vectors;

[0095] and These represent the total weight matrix. The submatrices corresponding to chemical and physical characteristics; The norm of a vector, such as the L2 norm;

[0096] This represents the chemical dominance contribution, calculated using S42, which quantifies the contribution of chemical factors deviating from the system's total potential energy gradient.

[0097] This represents the physical dominant contribution, calculated by S43, which quantifies the contribution of deviations from physical factors.

[0098] This represents the dominant pathway ratio, calculated from S44, and is a dimensionless scalar.

[0099] When the stability index When the warning threshold is broken, the system only knows that a failure is about to occur, but cannot distinguish whether the root cause of the failure is due to the chemical factors of the ink system or the physical factors of the equipment operation. To solve this ambiguity of cause, this embodiment proposes to decompose the total system potential energy gradient into two orthogonal components based on its physicochemical source, namely the chemical dominant contribution and the physical dominant contribution, so as to quantify the driving strength of each component on system instability.

[0100] when After the alarm is triggered, the system immediately performs parallel calculations. and And find the ratio between the two. By By comparing the result with a preset decision threshold (S45), the system can automatically determine the root cause; a value much greater than 1 A value of 1 indicates that the contribution of the chemical component is significantly greater than that of the physical component, and the system is judged to be chemically dominated failure; conversely, a value much less than 1 indicates a chemically dominated failure. The value is then determined to be a physical-dominated failure; this method directly solves the problem that traditional monitoring methods cannot distinguish whether the root cause of similar defects is abnormal ink viscosity in the chemical category or mechanical vibration in the physical category.

[0101] By statistically analyzing a large amount of historical data on the transition phase from normal operation to failure, different... The receiver operating characteristic curve (ROC curve) below the threshold is used to determine the receiver operating characteristic curve. On this curve, a point is selected that maximizes the alert recall rate while ensuring an acceptable false positive rate. This point corresponds to... The value is then determined as the warning threshold. ;

[0102] Scientific calibration was performed through controlled fault injection experiments to determine the chemically dominant decision threshold. In an experimental environment, a known problem-causing fast-evaporating solvent was artificially introduced into the ink system to actively induce chemical pathway failure, and the data under this condition was continuously recorded and analyzed. The statistical distribution is determined by selecting the lower bound of the distribution (e.g., the 5th percentile) as the statistical boundary. Similarly, to determine the threshold for physics-dominant decision-making... A known weak mechanical vibration excitation is applied to a key rotating component to simulate a physical path failure, and the results are recorded and analyzed. The statistical distribution is used, and its upper bound is selected as the statistical upper bound. This calibration method ensures the objectivity and robustness of the decision threshold.

[0103] Example 5:

[0104] According to the aforementioned method, S5 specifically includes: S51, if the dominant failure path is determined to be chemically dominant, an instruction is automatically generated and sent to the ink viscosity control system to adjust the solvent replenishment parameters, and at the same time, an abnormal ink status prompt is pushed to the human-machine interface; S52, if the dominant failure path is determined to be physically dominant, based on the characteristic frequency information in the vibration feature sub-vector, the specific abnormal vibration component is identified, and a high-priority preventive maintenance work order for the component is automatically generated.

[0105] The generation of differentiated intervention instructions in this embodiment is a closed-loop execution link in the entire detection and diagnosis method, directly reflecting the accuracy of diagnosis and the efficiency of intervention. Once the dominant failure pathway is determined, the precise differentiated intervention decision module in step S5 is activated. If the determination result of S51 is chemically dominant, for example, if a chemically dominant pathway is detected... Higher than The module will immediately generate control commands, automatically increasing the solvent replenishment pump frequency of the ink viscosity control system by a specific percentage via the industrial bus to address the identified problem of excessively rapid solvent evaporation. Simultaneously, it will push clear prompts to the operator's human-machine interface, such as "Ink drying speed is abnormal; the system has automatically increased the solvent replenishment rate. Please monitor subsequent quality changes." If the S52's determination is based on physical factors, such as detecting... Below The module will then conduct further in-depth analysis of the vibration feature sub-vectors. According to energy amplitude The largest characteristic frequency The system identifies specific sources of abnormal vibration, such as the No. 3 guide roller bearing, from a database mapping equipment components to characteristic frequencies. It then automatically generates a high-priority preventative maintenance work order through the manufacturing execution system. The work order clearly points to the faulty component, such as "Warning: Vibration of the No. 3 guide roller bearing exceeds limits. It is recommended to arrange an inspection or planned maintenance immediately." This intervention process is fully automated and completed within seconds, achieving precise, closed-loop, and proactive intelligent control. It effectively avoids production stoppages and material waste caused by human error, delayed intervention, or improper operation.

[0106] Example 6:

[0107] An automatic detection system for surface defects in printed materials using machine vision includes: a multimodal data synchronous acquisition module; a dynamic feature vector construction module; a system state stability evolution modeling module; a dominant failure path causal decoupling module; and a precise differentiated intervention decision module.

[0108] This embodiment also provides an automatic detection system for surface defects of printed materials that implements the above method; the system consists of the following functional modules:

[0109] The multimodal data synchronous acquisition module's hardware includes a high-resolution linear CCD camera, a piezoelectric accelerometer sensor installed at key locations on the device, an infrared thermal imager array, and a network time protocol server to ensure time consistency; this module is responsible for accurately and synchronously acquiring visual, vibration, and temperature data streams.

[0110] The dynamic feature vector construction module is typically materialized as an edge computing unit or embedded system. Internally, it integrates image processing algorithms (for calculating image blur, texture entropy, and color gradient norm), digital signal processing algorithms (for performing short-time Fourier transforms and extracting energy amplitude), and thermal imaging analysis algorithms. This allows it to transform massive amounts of raw sensor data into a unified and normalized system state vector in real time. ;

[0111] The system state stability evolution modeling module's core is a software package deployed on an industrial PC or server; this software module continuously receives... Vector flow, and based on a pre-calibrated reference steady-state vector. and weight matrix Real-time calculation of system potential energy and stability index This enables a quantitative assessment of the system's operating status and a prediction of its stable trends.

[0112] The primary failure path causal decoupling module, as a core logical component of the aforementioned software package, is responsible for decoupling the system when the system stability index falls below a preset warning threshold. When this module is conditionally triggered, it is responsible for performing the decomposition calculation of the dominant chemical and physical contributions, and based on the dominant pathway ratio. The comparison result with the decision threshold outputs a clear determination of the root cause of instability;

[0113] The precise differentiated intervention decision module serves as the final output and execution interface of the system. Based on the judgment conclusion of the decoupling module, this module sends control commands directly to the programmable logic controller of the ink viscosity control system via the industrial fieldbus protocol, or pushes electronic preventive maintenance work orders containing precise fault location information to the management terminal of the equipment maintenance department via the enterprise resource planning system.

[0114] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An automatic detection method for surface defects in printed materials using machine vision, characterized in that, Includes the following steps: S1. Synchronously collect time-series defect image data, key component vibration data, and substrate surface temperature data during the printing process to form a multimodal time-series dataset; S2. Extract and quantify key dynamic features from the multimodal time-series dataset to generate a unified system state vector; the key dynamic features include visual features, vibrational features, and thermodynamic features; S3. Construct a system state stability evolution model, input the system state vector into the stability evolution model, calculate and output the real-time system stability index; the stability evolution model is constructed based on the deviation between the system state vector and the preset benchmark stable state vector; S4. In response to the real-time system stability index falling below a preset warning threshold, initiate causal decoupling analysis, calculate the chemical dominant contribution and physical dominant contribution based on the system state vector, and determine the current dominant failure path by comparing the chemical dominant contribution and physical dominant contribution. S5. Based on the dominant failure path, generate differentiated intervention instructions; if the dominant failure path is determined to be chemically dominant, generate adjustment instructions for the ink mixing system. If the dominant failure path is determined to be physically dominant, an early warning instruction for the equipment maintenance system is generated. S3 specifically includes: S31. Based on the system state vector and the preset benchmark stable state vector, calculate the difference between the two to obtain the system state deviation vector; S32. By performing a weighted quadratic form calculation on the system state deviation vector using a preset weight matrix, a system potential energy value is generated that characterizes the degree to which the system deviates from the stable point in the potential energy field. S33. Based on the time series of the system potential energy value, the real-time system stability index is obtained by calculating the product of the autocorrelation and variance of the time series of the system potential energy value and taking the reciprocal of the product. The baseline steady-state vector is obtained by continuously collecting and analyzing printing process data with a preset yield rate higher than a specific standard, and by calculating the time average of the calculated system state vector at each moment. The preset weight matrix is ​​optimized by using Lasso regression analysis to analyze historical fault data and quantify the contribution of each key dynamic feature to the final product quality defects. S4 specifically includes: S41. Decouple the system state vector and the preset weight matrix into a chemical part and a physical part, respectively. The chemical part corresponds to the visual and thermodynamic characteristics related to ink, and the physical part corresponds to the vibration characteristics related to equipment. S42. Calculate the chemical dominant contribution based on the system state vector, baseline stable state vector, and weight matrix of the chemical component; S43. Calculate the physical dominant contribution based on the system state vector, baseline stable state vector, and weight matrix of the physical component; S44. Calculate the ratio of the chemical dominant contribution to the physical dominant contribution to obtain the dominant pathway ratio; S45. If the dominant pathway ratio is higher than a preset chemical dominance decision threshold, the dominant failure pathway is determined to be chemically dominant; if the dominant pathway ratio is lower than a preset physical dominance decision threshold, the dominant failure pathway is determined to be physically dominant.

2. The automatic detection method for surface defects of printed materials using machine vision according to claim 1, characterized in that, S2 specifically includes: S21. Based on the temporal defect image data, visual feature sub-vectors are extracted and combined by calculating image blur, texture entropy and color gradient norm. S22. Based on the vibration data of key components, the energy amplitude at the relevant characteristic frequencies of the key components is extracted by performing a short-time Fourier transform on the vibration data of the key components, and combined into a vibration feature sub-vector. S23. Based on the surface temperature data of the substrate, the average surface temperature, temperature standard deviation and maximum surface temperature gradient are calculated, and thermodynamic feature sub-vectors are extracted and combined. S24. Normalize and merge the visual feature sub-vector, the vibration feature sub-vector, and the thermodynamic feature sub-vector into the unified system state vector.

3. The automatic detection method for surface defects of printed materials using machine vision according to claim 2, characterized in that, The chemically dominant decision threshold and the physically dominant decision threshold are obtained through controlled fault injection experiments, specifically including: a) By introducing a disturbance of excessively rapid solvent evaporation into the ink system during the experiment, the statistical distribution of the dominant pathway ratio under this condition was continuously recorded and analyzed to determine the chemical dominance decision threshold; b) By applying known weak mechanical vibration excitation to key rotating components in experiments, continuously recording and analyzing the statistical distribution of the dominant pathway ratio under this condition, the physical dominance decision threshold can be determined.

4. The automatic detection method for surface defects of printed materials using machine vision according to claim 1, characterized in that, The preset warning threshold is determined by statistically analyzing a large amount of system data transitioning from a normal state to a fault state, and based on the receiver's operational characteristic curve analysis, selecting the corresponding value that maximizes the warning recall rate while ensuring an acceptable false positive rate.

5. The automatic detection method for surface defects of printed materials using machine vision according to claim 1, characterized in that, S5 specifically includes: S51. If the dominant failure pathway is determined to be chemically dominant, an instruction will be automatically generated and sent to the ink viscosity control system to adjust the solvent replenishment parameters, and at the same time, an abnormal ink status prompt message will be pushed to the human-machine interface. S52. If the dominant failure path is determined to be physically dominant, then based on the characteristic frequency information in the vibration characteristic sub-vector, the specific abnormal vibration component is identified, and a high-priority preventive maintenance work order for that component is automatically generated.

6. An automatic detection system for surface defects of printed materials using machine vision, employing the automatic detection method for surface defects of printed materials using machine vision as described in any one of claims 1-5, characterized in that, include: The multimodal data synchronous acquisition module is used to synchronously acquire temporal defect image data, key component vibration data, and substrate surface temperature data during the printing process to form a multimodal time-series dataset. The dynamic feature vector construction module is used to extract and quantify key dynamic features reflecting the system state from the multimodal time series dataset and fuse them into a unified system state vector. The system state stability evolution modeling module is used to construct a model describing the dynamic evolution of the system state vector based on the system state vector, and to calculate and output the real-time system stability index based on the model. The dominant failure path causal decoupling module is activated in response to a real-time system stability index falling below a preset warning threshold to analyze the system state vector, thereby decoupling and determining the chemically or physically dominant failure path that causes system instability. The precise differentiated intervention decision module is used to generate and output differentiated control instructions for chemical or physical pathways based on the dominant failure pathway determined by the causal decoupling module.

Images